Hybrid Search with Mistral

This recipe will show you how to run hybrid search with embeddings from Mistral.

Requirements

Weaviate cluster
1. You can create a free cluster on WCD
2. Embedded Weaviate
3. Local deployment
4. Other options
Mistral API key. Grab one here.

Python

import weaviate, os
from weaviate.embedded import EmbeddedOptions
import weaviate.classes as wvc
import weaviate.classes.config as wc
import requests, json
import weaviate.classes.query as wq

Connect to Weaviate

Only choose one option from the below.

Weaviate Cloud Deployment

Python

WCD_URL = os.environ["WEAVIATE_URL"] # Replace with your Weaviate cluster URL
WCD_AUTH_KEY = os.environ["WEAVIATE_AUTH"] # Replace with your cluster auth key
MISTRAL_KEY = os.environ["MISTRAL_API_KEY"] # Replace with your Mistral key

# Weaviate Cloud Deployment
client = weaviate.connect_to_wcs(
    cluster_url=WCD_URL,
    auth_credentials=weaviate.auth.AuthApiKey(WCD_AUTH_KEY),
      headers={ "X-Mistral-Api-Key": MISTRAL_KEY}
)

print(client.is_ready())

Embedded Weaviate

Python

# MISTRAL_KEY = os.environ["MISTRAL_API_KEY"] # Replace with your Mistral key

# client = weaviate.WeaviateClient(
#     embedded_options=EmbeddedOptions(
#         version="1.26.1",
#         additional_env_vars={
#             "ENABLE_MODULES": "text2vec-mistral"
#         }),
#         additional_headers={
#             "X-Mistral-Api-Key": MISTRAL_KEY
#         }
# )

# client.connect()

Local Deployment

Python

# MISTRAL_KEY = os.environ["MISTRAL_API_KEY"] # Replace with your Mistral key

# client = weaviate.connect_to_local(
#   headers={
#     "X-Mistral-Api-Key": MISTRAL_KEY
#   }
# )
# print(client.is_ready())

Create a collection

Collection stores your data and vector embeddings.

Python

# Note: in practice, you shouldn't rerun this cell, as it deletes your data
# in "JeopardyQuestion", and then you need to re-import it again.

# Delete the collection if it already exists
if (client.collections.exists("JeopardyQuestion")):
    client.collections.delete("JeopardyQuestion")

client.collections.create(
    name="JeopardyQuestion",

    vector_config=wc.Configure.Vectors.text2vec_mistral( # specify the vectorizer and model
        model="mistral-embed",
    ),

    properties=[ # defining properties (data schema) is optional
        wc.Property(name="Question", data_type=wc.DataType.TEXT),
        wc.Property(name="Answer", data_type=wc.DataType.TEXT),
        wc.Property(name="Category", data_type=wc.DataType.TEXT, skip_vectorization=True),
    ]
)

print("Successfully created collection: JeopardyQuestion.")

Import Data

Python

url = 'https://raw.githubusercontent.com/weaviate/weaviate-examples/main/jeopardy_small_dataset/jeopardy_tiny.json'
resp = requests.get(url)
data = json.loads(resp.text)

# Get a collection object for "JeopardyQuestion"
jeopardy = client.collections.use("JeopardyQuestion")

# Insert data objects
response = jeopardy.data.insert_many(data)

# Note, the `data` array contains 10 objects, which is great to call insert_many with.
# However, if you have a milion objects to insert, then you should spit them into smaller batches (i.e. 100-1000 per insert)

if (response.has_errors):
    print(response.errors)
else:
    print("Insert complete.")

Hybrid Search

The alpha parameter determines the weight given to the sparse and dense search methods. alpha = 0 is pure sparse (bm25) search, whereas alpha = 1 is pure dense (vector) search.

Alpha is an optional parameter. The default is set to 0.75.

Hybrid Search only

The below query is finding Jeopardy questions about animals and is limiting the output to only two results. Notice alpha is set to 0.80, which means it is weighing the vector search results more than bm25. If you were to set alpha = 0.25, you would get different results.

Python

response = jeopardy.query.hybrid(
    query="northern beast",
    query_properties=["question"],
    alpha=0.8,
    limit=3
)

for item in response.objects:
    print("ID:", item.uuid)
    print("Data:", json.dumps(item.properties, indent=2), "\n")

Hybrid Search with a `where` filter

Find Jeopardy questions about elephants, where the category is set to Animals.

Python

response = jeopardy.query.hybrid(
    query="northern beast",
    alpha=0.8,
    filters=wq.Filter.by_property("category").equal("Animals"),
    limit=3
)

for item in response.objects:
    print("ID:", item.uuid)
    print("Data:", json.dumps(item.properties, indent=2), "\n")

Hybrid Search with Mistral

Requirements

Connect to Weaviate

Create a collection

Import Data

Hybrid Search

Hybrid Search only

Hybrid Search with a where filter

Hybrid Search with a `where` filter